McKinsey-Style Consulting Deliverable Generator

pip install requests python-dotenv
export SWARMS_API_KEY="your-api-key-here"

import json
import os

import requests
from dotenv import load_dotenv

load_dotenv()

API_KEY = os.getenv("SWARMS_API_KEY")
BASE_URL = "https://api.swarms.world"

headers = {"x-api-key": API_KEY, "Content-Type": "application/json"}

def build_engagement(client_brief: str) -> dict:
    return {
        "name": "Consulting-Deliverable-Generator",
        "description": (
            "Hierarchical engagement: Engagement Director coordinates four "
            "specialists and produces a partner-ready slide deck outline."
        ),
        "swarm_type": "HierarchicalSwarm",
        "task": client_brief,
        "max_loops": 1,
        "agents": [
            {
                "agent_name": "Engagement Director",
                "description": "Senior partner-level lead. Synthesizes all worker outputs into a slide-by-slide deck.",
                "system_prompt": (
                    "You are the Engagement Director on a market-entry study. You have "
                    "four specialists reporting to you: an Industry Analyst, a Customer "
                    "Researcher, a Competitor Analyst, and a Financial Modeler.\n\n"
                    "Your job is NOT to write a memo. Your job is to produce a "
                    "partner-ready slide-by-slide deck outline using the Pyramid "
                    "Principle (answer first, then supporting structure). Use this "
                    "exact format for every slide:\n\n"
                    "SLIDE N: <title>\n"
                    "  Headline (governing thought, one sentence):\n"
                    "  Body (3-5 bullets, MECE):\n"
                    "  Chart / exhibit callout: <what visual goes here>\n"
                    "  Source: <which specialist(s) supplied this>\n\n"
                    "Required slides, in order:\n"
                    "  1. Title slide\n"
                    "  2. Executive summary (the answer)\n"
                    "  3. Context & client question\n"
                    "  4. Market sizing (TAM/SAM/SOM)\n"
                    "  5. Customer landscape & unmet need\n"
                    "  6. Competitor landscape (2x2 or matrix)\n"
                    "  7. Unit economics & financial model summary\n"
                    "  8. Strategic options considered\n"
                    "  9. Recommendation\n"
                    " 10. Risks & mitigations\n"
                    " 11. 90-day implementation roadmap\n"
                    " 12. Appendix index\n\n"
                    "The Recommendation slide must give a clear ENTER / ENTER-PARTNERED / "
                    "WAIT / PASS verdict with the two strongest supporting reasons and "
                    "the single largest risk."
                ),
                "model_name": "gpt-4.1",
                "role": "coordinator",
                "max_loops": 1,
                "max_tokens": 8192,
                "temperature": 0.3,
            },
            {
                "agent_name": "Industry Analyst",
                "description": "Sizes the market and profiles industry structure.",
                "system_prompt": (
                    "You are a senior industry analyst. For the client's target market: "
                    "estimate TAM, SAM, and SOM with named assumptions; describe value "
                    "chain and economics; identify regulatory regime; flag the three "
                    "structural trends most likely to move the market over the next "
                    "five years. Cite the type of source you would pull each number "
                    "from (e.g., 'FDIC Call Reports', 'SBA loan data') even if you "
                    "estimate from prior knowledge. Be quantitative."
                ),
                "model_name": "gpt-4.1",
                "role": "worker",
                "max_loops": 1,
                "max_tokens": 4096,
                "temperature": 0.4,
            },
            {
                "agent_name": "Customer Researcher",
                "description": "Profiles target customers, segments, and unmet needs.",
                "system_prompt": (
                    "You are a senior customer research lead. Segment the target "
                    "customer base into 3-4 archetypes with named attributes (size, "
                    "vertical, sophistication, geography). For each segment: estimate "
                    "size, willingness to pay, top three jobs-to-be-done, and the "
                    "single largest unmet need today. Identify which segment is the "
                    "best beachhead and why."
                ),
                "model_name": "gpt-4.1",
                "role": "worker",
                "max_loops": 1,
                "max_tokens": 4096,
                "temperature": 0.4,
            },
            {
                "agent_name": "Competitor Analyst",
                "description": "Maps the competitive landscape and identifies whitespace.",
                "system_prompt": (
                    "You are a senior competitor analyst. Identify the top 5-8 "
                    "competitors the client will face, including incumbents and "
                    "challengers. For each: business model, served segment, "
                    "estimated share, key strength, key weakness. Position them on "
                    "a 2x2 of your choosing (state the axes explicitly) and "
                    "identify the whitespace the client could occupy. Call out any "
                    "competitor whose response could realistically kill the entry."
                ),
                "model_name": "gpt-4.1",
                "role": "worker",
                "max_loops": 1,
                "max_tokens": 4096,
                "temperature": 0.3,
            },
            {
                "agent_name": "Financial Modeler",
                "description": "Builds the unit-economics and three-year P&L sketch.",
                "system_prompt": (
                    "You are a senior financial modeler. Produce a defensible "
                    "first-cut model: unit economics (revenue per customer, gross "
                    "margin, CAC, payback period), three-year P&L sketch with stated "
                    "assumptions, capital required to reach breakeven, and the two "
                    "or three assumptions that most move the answer. Identify the "
                    "single most important sensitivity and state which way it cuts."
                ),
                "model_name": "gpt-4.1",
                "role": "worker",
                "max_loops": 1,
                "max_tokens": 4096,
                "temperature": 0.3,
            },
        ],
    }

client_brief = (
    "We are a mid-market regional bank with $14B in assets, headquartered in "
    "the Southeast U.S. Our retail and CRE books are mature. Our board is "
    "asking us to evaluate whether to enter SMB lending (loans of $50k-$2M to "
    "businesses with under $25M revenue) as a third growth pillar. Build us a "
    "market-entry analysis: should we enter, how, and what would it take?"
)

payload = build_engagement(client_brief)

response = requests.post(
    f"{BASE_URL}/v1/swarm/completions",
    headers=headers,
    json=payload,
    timeout=300,
)

result = response.json()

for output in result.get("output", []):
    print("=" * 60)
    print(output["role"])
    print("=" * 60)
    content = output["content"]
    if isinstance(content, list):
        content = " ".join(str(c) for c in content)
    print(str(content)[:600] + "...\n")

usage = result.get("usage", {})
billing = usage.get("billing_info", {})
print(f"\nTotal cost: ${billing.get('total_cost', 0):.4f}")
print(f"Execution time: {result.get('execution_time', 'n/a')}s")

client_briefs = [
    "Mid-market regional bank evaluating SMB lending entry — see prior brief.",
    "$3B AUM PE firm evaluating a roll-up in U.S. veterinary clinics.",
    "European industrial OEM evaluating direct-to-consumer entry in North America.",
    # ...add as many as you need
]

batch_payload = [build_engagement(b) for b in client_briefs]

batch_response = requests.post(
    f"{BASE_URL}/v1/swarm/batch/completions",
    headers=headers,
    json=batch_payload,
    timeout=1800,
)

decks = batch_response.json()
print(f"Decks produced: {len(decks)}")